This proposal presents four mathematical models of retrieval processes involved in cued recall of responses from well-memorized lists of paired-associates. The models differ in their treatment of such issues as whether retrieval of a response always involves a scan of memory, and whether the scan of memory has its locus in long or short-term storage. We propose to examine reaction time as a function of list length and of sequencing of the probes. Our earlier studies are exploratory, permitting us to estimate parameters of our models and then fit various statistics of the data. Subsequently, we will carry out some experiments that bear directly on such issues as whether a scan takes place, and where. Finally, we propose experiments in which semantic variables are manipulated. The results provide a basis for elaborating on storage mechanisms in retrieval within the context of our models.